Re: [PATCH 3/8] acpi-cpufreq: merge acpi functionality of cpufreq drivers

[Bruno Ducrot <ducrot@poupinou.org>]

On Fri, Aug 04, 2006 at 01:03:38PM -0400, Dave Jones wrote:
> On Fri, Aug 04, 2006 at 11:22:12AM +0200, Bruno Ducrot wrote:
> 
>  > > Newer VIA CPUs rely on this driver rather than longhaul for example.
>  > 
>  > Ah I see.  In fact we could add them already to
>  > speedstep-centrino.c easily.
> 
> We could, though they need to rely on the ACPI tables rather than
> hardcoded tables (as we've no way to distinguish regular from ULV
> parts afaics), at which point it doesn't really buy us anything
> over just using acpi-cpufreq which works without any modifications.

I don't get it.  speedstep-centrino does support both hardcoded tables
and ACPI tables (but I agree it would be better to add this support
to acpi-cpufreq instead of speedstep-centrino).

FYI FreeBSD support VIA C7-M processors with hardcoded tables.  They
use MSR_IA32_PERF_STATUS, bit 31 to 64, in order to distinguish various
variants of those processors (and Intel processors as well btw).
See /sys/i386/cpufreq/est.c
http://www.freebsd.org/cgi/cvsweb.cgi/~checkout~/src/sys/i386/cpufreq/est.c?rev=1.11&content-type=text/plain

-- 
Bruno Ducrot

--  Which is worse:  ignorance or apathy?
--  Don't know.  Don't care.